“The purpose of this collaboration, still in an exploratory phase, has been to determine what weights in an investment portfolio containing certain assets yield higher returns and, in financial terms, we have come up with a way to optimize these calculations that hadn’t been considered until now. Methods based on quantum computing and quantum-inspired methods are new and can improve the current tools that are already in practice,” explains BBVA head of Research and Patents discipline Escolástico Sánchez. Sánchez is one of the authors of this paper together with doctors Román Orús, Enrique Lizaso and Sam Mugel, from Multiverse Computing; as well as BBVA global head of Research and Patents, Carlos Kuchkovsky; and Jorge Luis Hita and Samuel Fernández, members of BBVA’s quantum algorithms team.
As Dr. Román Orús, co-founder and CSO at Multiverse Computing explained, “we have implemented the optimization with real market data for the first time applying a variational algorithm for IBM-Q, a hybrid quantum-classical algorithm for D-Wave and, for the first time internationally in a financial context, using Tensor Networks’ ‘quantum-inspired’ algorithms ”.
Multiverse is a Spanish tech startup that specializes in developing quantum algorithms for the international financial sector. The company is run by a prestigious team of experts in quantum physics, artificial intelligence, machine learning, mathematics and economics. Since it launched, it has garnered support from tech accelerators and hubs in the Basque Country such as the Donostia International Physics Center and BIC Gipuzkoa, as well as Canada’s Creative Destruction Lab.
A classic problem in finance
When setting up an investment portfolio, the key focus is finding the right combination of assets to maximize returns while minimizing risk. Doing so requires taking into account many factors that can affect asset performance and which vary over time. One way to approach this task is configuring portfolios dynamically, that is, varying the weight of the different assets that make up the portfolio periodically over time. Defining the optimal path of these values considering all the factors that affect how they fluctuate – such as transaction fee costs – is a well-known problem in the world of finance which cannot be tackled using traditional techniques.
In the proof of concept, researchers have attempted to solve this problem resorting to a combination of quantum technology platforms, inspired by quantum computing and traditional techniques, and compared the results. Their conclusions suggest that quantum computing-based tools and quantum-inspired algorithms, could already perform this task more efficiently than traditional methods. Furthermore, according to the authors, this is the first time that quantum technology tools have been leveraged to optimize an investment portfolio large enough to carry commercial value.
Specifically, the tests were run on different ‘hardware’ platforms in which a series of quantum and quantum-inspired algorithms had been implemented. The idea was to determine the optimal trading path for an investment portfolio consisting of 52 assets, using actual daily market price data corresponding to an 8-year timeframe. Finding the best investment portfolio among a sample of 10,382 candidates and doing it dynamically, i.e. in such a way that portfolio weights fluctuated depending on the purchases and sales of the market. Processing such a large amount of data would have taken a traditional computer using conventional algorithms approximately two days. On the contrary, according to the results of his work, quantum algorithms should be capable of performing them immediately, in a matter of seconds.
Four new and two classic methods
To carry out these tests, the authors compared the ‘Sharpe ratio’ – a ratio used to measure the return of an investment compared to its risk –, profit returns and computation times obtained by performing the task using the different technological solutions.
They assessed and compared four quantum computing-based methods and two classical methods used in finance. Regarding quantum technologies, a method based on hybrid computing was run on the D-Wave Hybrid, as well as two other approaches built on the VQE (Variational Quantum Eigensolvers) algorithm, deployed in IBM’s quantum computer Q System One. Finally, an attempt was made to solve the problem using a quantum-inspired algorithm based on Tensor Networks’ platform.
In the case of the quantum-inspired algorithm can be used in conventional computers, the authors believe that it offered promising results and that it was the first time it had ever been applied to solve this problem.
A leap forward
The test carried out between BBVA and Multiverse places the bank at the cutting edge in the deployment of quantum technologies applied to quantitative finance. Since 2018, BBVA has collaborated in a number of quantum computing projects partnering with companies such as Fujitsu, Accenture, U.S. startup Zapata and Spain’s Search Results Spanish National Research Council), with which it signed a collaboration partnership in 2019.
One of its researchers at the Institute of Fundamental Physics, Juan José García Ripoll, defines the collaboration between Multiverse and BBVA as follows: “Philosophy is to take a specific problem and try to identify quantum and classical ways to solve it. The good thing about this process is that as we learn how quantum computers work, we also discover improvements or alternatives to classic methods that are better than the ones we have available.” Through their partnership, BBVA and CSIC collaborate on a line of research targeted at developing its own quantum algorithms.
Although quantum computing is still in its infant stages and machines are still far from flawless, the sector and research have been growing at substantial rates in recent years. BBVA’s interest in this area is aligned with its goal to look into disruptive technologies and trends with significant disruptive potential for the financial sector, setting up cross-disciplinary teams specialized in advanced technological and scientific disciplines. “This field is growing fast because the limitations imposed by the shortage of machines and computers have disappeared. That has increased the capacity. The challenge now is to keep building and sustain growth by nurturing the sector, the labs and the startups led by up-and-coming talent so that they can focus on developing the hardware and software infrastructure,” highlights García Ripoll.
As Dr. Román Orús, co-founder and CSO at Multiverse Computing explained, “we have implemented the optimization with real market data for the first time applying a variational algorithm for IBM-Q, a hybrid quantum-classical algorithm for D-Wave and, for the first time internationally in a financial context, using Tensor Networks’ ‘quantum-inspired’ algorithms ”.
Multiverse is a Spanish tech startup that specializes in developing quantum algorithms for the international financial sector. The company is run by a prestigious team of experts in quantum physics, artificial intelligence, machine learning, mathematics and economics. Since it launched, it has garnered support from tech accelerators and hubs in the Basque Country such as the Donostia International Physics Center and BIC Gipuzkoa, as well as Canada’s Creative Destruction Lab.
A classic problem in finance
When setting up an investment portfolio, the key focus is finding the right combination of assets to maximize returns while minimizing risk. Doing so requires taking into account many factors that can affect asset performance and which vary over time. One way to approach this task is configuring portfolios dynamically, that is, varying the weight of the different assets that make up the portfolio periodically over time. Defining the optimal path of these values considering all the factors that affect how they fluctuate – such as transaction fee costs – is a well-known problem in the world of finance which cannot be tackled using traditional techniques.
In the proof of concept, researchers have attempted to solve this problem resorting to a combination of quantum technology platforms, inspired by quantum computing and traditional techniques, and compared the results. Their conclusions suggest that quantum computing-based tools and quantum-inspired algorithms, could already perform this task more efficiently than traditional methods. Furthermore, according to the authors, this is the first time that quantum technology tools have been leveraged to optimize an investment portfolio large enough to carry commercial value.
Specifically, the tests were run on different ‘hardware’ platforms in which a series of quantum and quantum-inspired algorithms had been implemented. The idea was to determine the optimal trading path for an investment portfolio consisting of 52 assets, using actual daily market price data corresponding to an 8-year timeframe. Finding the best investment portfolio among a sample of 10,382 candidates and doing it dynamically, i.e. in such a way that portfolio weights fluctuated depending on the purchases and sales of the market. Processing such a large amount of data would have taken a traditional computer using conventional algorithms approximately two days. On the contrary, according to the results of his work, quantum algorithms should be capable of performing them immediately, in a matter of seconds.
Four new and two classic methods
To carry out these tests, the authors compared the ‘Sharpe ratio’ – a ratio used to measure the return of an investment compared to its risk –, profit returns and computation times obtained by performing the task using the different technological solutions.
They assessed and compared four quantum computing-based methods and two classical methods used in finance. Regarding quantum technologies, a method based on hybrid computing was run on the D-Wave Hybrid, as well as two other approaches built on the VQE (Variational Quantum Eigensolvers) algorithm, deployed in IBM’s quantum computer Q System One. Finally, an attempt was made to solve the problem using a quantum-inspired algorithm based on Tensor Networks’ platform.
In the case of the quantum-inspired algorithm can be used in conventional computers, the authors believe that it offered promising results and that it was the first time it had ever been applied to solve this problem.
A leap forward
The test carried out between BBVA and Multiverse places the bank at the cutting edge in the deployment of quantum technologies applied to quantitative finance. Since 2018, BBVA has collaborated in a number of quantum computing projects partnering with companies such as Fujitsu, Accenture, U.S. startup Zapata and Spain’s Search Results Spanish National Research Council), with which it signed a collaboration partnership in 2019.
One of its researchers at the Institute of Fundamental Physics, Juan José García Ripoll, defines the collaboration between Multiverse and BBVA as follows: “Philosophy is to take a specific problem and try to identify quantum and classical ways to solve it. The good thing about this process is that as we learn how quantum computers work, we also discover improvements or alternatives to classic methods that are better than the ones we have available.” Through their partnership, BBVA and CSIC collaborate on a line of research targeted at developing its own quantum algorithms.
Although quantum computing is still in its infant stages and machines are still far from flawless, the sector and research have been growing at substantial rates in recent years. BBVA’s interest in this area is aligned with its goal to look into disruptive technologies and trends with significant disruptive potential for the financial sector, setting up cross-disciplinary teams specialized in advanced technological and scientific disciplines. “This field is growing fast because the limitations imposed by the shortage of machines and computers have disappeared. That has increased the capacity. The challenge now is to keep building and sustain growth by nurturing the sector, the labs and the startups led by up-and-coming talent so that they can focus on developing the hardware and software infrastructure,” highlights García Ripoll.